X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/195f9f5ef3f79ea4c77ee1226f8aa7e7e8f4a73f..70dc4957dc41df7d808ae4acb6c881e80b6348f9:/nominatim/clicmd/refresh.py?ds=inline diff --git a/nominatim/clicmd/refresh.py b/nominatim/clicmd/refresh.py index 4e2ad1db..ea605ea0 100644 --- a/nominatim/clicmd/refresh.py +++ b/nominatim/clicmd/refresh.py @@ -1,11 +1,21 @@ +# SPDX-License-Identifier: GPL-2.0-only +# +# This file is part of Nominatim. (https://nominatim.org) +# +# Copyright (C) 2022 by the Nominatim developer community. +# For a full list of authors see the git log. """ Implementation of 'refresh' subcommand. """ +from typing import Tuple, Optional +import argparse import logging from pathlib import Path -from ..db.connection import connect -from ..tools.exec_utils import run_legacy_script +from nominatim.config import Configuration +from nominatim.db.connection import connect +from nominatim.tokenizer.base import AbstractTokenizer +from nominatim.clicmd.args import NominatimArgs # Do not repeat documentation of subcommand classes. # pylint: disable=C0111 @@ -14,18 +24,37 @@ from ..tools.exec_utils import run_legacy_script LOG = logging.getLogger() +def _parse_osm_object(obj: str) -> Tuple[str, int]: + """ Parse the given argument into a tuple of OSM type and ID. + Raises an ArgumentError if the format is not recognized. + """ + if len(obj) < 2 or obj[0].lower() not in 'nrw' or not obj[1:].isdigit(): + raise argparse.ArgumentTypeError("Cannot parse OSM ID. Expect format: [N|W|R].") + + return (obj[0].upper(), int(obj[1:])) + + class UpdateRefresh: """\ Recompute auxiliary data used by the indexing process. - These functions must not be run in parallel with other update commands. + This sub-commands updates various static data and functions in the database. + It usually needs to be run after changing various aspects of the + configuration. The configuration documentation will mention the exact + command to use in such case. + + Warning: the 'update' command must not be run in parallel with other update + commands like 'replication' or 'add-data'. """ + def __init__(self) -> None: + self.tokenizer: Optional[AbstractTokenizer] = None - @staticmethod - def add_args(parser): + def add_args(self, parser: argparse.ArgumentParser) -> None: group = parser.add_argument_group('Data arguments') group.add_argument('--postcodes', action='store_true', help='Update postcode centroid table') + group.add_argument('--word-tokens', action='store_true', + help='Clean up search terms') group.add_argument('--word-counts', action='store_true', help='Compute frequency of full-word search terms') group.add_argument('--address-levels', action='store_true', @@ -33,56 +62,122 @@ class UpdateRefresh: group.add_argument('--functions', action='store_true', help='Update the PL/pgSQL functions in the database') group.add_argument('--wiki-data', action='store_true', - help='Update Wikipedia/data importance numbers.') + help='Update Wikipedia/data importance numbers') + group.add_argument('--secondary-importance', action='store_true', + help='Update secondary importance raster data') group.add_argument('--importance', action='store_true', help='Recompute place importances (expensive!)') group.add_argument('--website', action='store_true', help='Refresh the directory that serves the scripts for the web API') + group.add_argument('--data-object', action='append', + type=_parse_osm_object, metavar='OBJECT', + help='Mark the given OSM object as requiring an update' + ' (format: [NWR])') + group.add_argument('--data-area', action='append', + type=_parse_osm_object, metavar='OBJECT', + help='Mark the area around the given OSM object as requiring an update' + ' (format: [NWR])') + group = parser.add_argument_group('Arguments for function refresh') group.add_argument('--no-diff-updates', action='store_false', dest='diffs', help='Do not enable code for propagating updates') group.add_argument('--enable-debug-statements', action='store_true', help='Enable debug warning statements in functions') - @staticmethod - def run(args): - from ..tools import refresh + + def run(self, args: NominatimArgs) -> int: #pylint: disable=too-many-branches, too-many-statements + from ..tools import refresh, postcodes + from ..indexer.indexer import Indexer + if args.postcodes: - LOG.warning("Update postcodes centroid") - conn = connect(args.config.get_libpq_dsn()) - refresh.update_postcodes(conn, args.data_dir) - conn.close() + if postcodes.can_compute(args.config.get_libpq_dsn()): + LOG.warning("Update postcodes centroid") + tokenizer = self._get_tokenizer(args.config) + postcodes.update_postcodes(args.config.get_libpq_dsn(), + args.project_dir, tokenizer) + indexer = Indexer(args.config.get_libpq_dsn(), tokenizer, + args.threads or 1) + indexer.index_postcodes() + else: + LOG.error("The place table doesn't exist. " + "Postcode updates on a frozen database is not possible.") + + if args.word_tokens: + LOG.warning('Updating word tokens') + tokenizer = self._get_tokenizer(args.config) + tokenizer.update_word_tokens() if args.word_counts: - LOG.warning('Recompute frequency of full-word search terms') - conn = connect(args.config.get_libpq_dsn()) - refresh.recompute_word_counts(conn, args.data_dir) - conn.close() + LOG.warning('Recompute word statistics') + self._get_tokenizer(args.config).update_statistics() if args.address_levels: - cfg = Path(args.config.ADDRESS_LEVEL_CONFIG) - LOG.warning('Updating address levels from %s', cfg) - conn = connect(args.config.get_libpq_dsn()) - refresh.load_address_levels_from_file(conn, cfg) - conn.close() + LOG.warning('Updating address levels') + with connect(args.config.get_libpq_dsn()) as conn: + refresh.load_address_levels_from_config(conn, args.config) + + # Attention: must come BEFORE functions + if args.secondary_importance: + with connect(args.config.get_libpq_dsn()) as conn: + # If the table did not exist before, then the importance code + # needs to be enabled. + if not conn.table_exists('secondary_importance'): + args.functions = True + + LOG.warning('Import secondary importance raster data from %s', args.project_dir) + if refresh.import_secondary_importance(args.config.get_libpq_dsn(), + args.project_dir) > 0: + LOG.fatal('FATAL: Cannot update sendary importance raster data') + return 1 if args.functions: LOG.warning('Create functions') - conn = connect(args.config.get_libpq_dsn()) - refresh.create_functions(conn, args.config, args.data_dir, - args.diffs, args.enable_debug_statements) - conn.close() + with connect(args.config.get_libpq_dsn()) as conn: + refresh.create_functions(conn, args.config, + args.diffs, args.enable_debug_statements) + self._get_tokenizer(args.config).update_sql_functions(args.config) if args.wiki_data: - run_legacy_script('setup.php', '--import-wikipedia-articles', - nominatim_env=args, throw_on_fail=True) + data_path = Path(args.config.WIKIPEDIA_DATA_PATH + or args.project_dir) + LOG.warning('Import wikipdia article importance from %s', data_path) + if refresh.import_wikipedia_articles(args.config.get_libpq_dsn(), + data_path) > 0: + LOG.fatal('FATAL: Wikipedia importance dump file not found') + return 1 + # Attention: importance MUST come after wiki data import. if args.importance: - run_legacy_script('update.php', '--recompute-importance', - nominatim_env=args, throw_on_fail=True) + LOG.warning('Update importance values for database') + with connect(args.config.get_libpq_dsn()) as conn: + refresh.recompute_importance(conn) + if args.website: - run_legacy_script('setup.php', '--setup-website', - nominatim_env=args, throw_on_fail=True) + webdir = args.project_dir / 'website' + LOG.warning('Setting up website directory at %s', webdir) + # This is a little bit hacky: call the tokenizer setup, so that + # the tokenizer directory gets repopulated as well, in case it + # wasn't there yet. + self._get_tokenizer(args.config) + with connect(args.config.get_libpq_dsn()) as conn: + refresh.setup_website(webdir, args.config, conn) + + if args.data_object or args.data_area: + with connect(args.config.get_libpq_dsn()) as conn: + for obj in args.data_object or []: + refresh.invalidate_osm_object(*obj, conn, recursive=False) + for obj in args.data_area or []: + refresh.invalidate_osm_object(*obj, conn, recursive=True) + conn.commit() return 0 + + + def _get_tokenizer(self, config: Configuration) -> AbstractTokenizer: + if self.tokenizer is None: + from ..tokenizer import factory as tokenizer_factory + + self.tokenizer = tokenizer_factory.get_tokenizer_for_db(config) + + return self.tokenizer